Unifying Vascular Information in Intensity-Based Nonrigid Lung CT Registration

摘要: Image registration plays an important role within pulmonary image analysis. Accurate is critical to post-analysis of lung mechanical properties. To improve accuracy, we utilize the rich information vessel locations and shapes, introduce a new similarity criterion, sum squared vesselness measure difference (SSVMD). This metric added three existing intensity-based criteria for nonrigid CT show its ability in improving matching accuracy. The accuracy assessed by landmark error calculation distance map visualization on vascular tree. average errors are reduced over 20% 0.7 mm after adding SSVMD constraint metrics. Visual inspection shows improvements regions near thoracic cage diaphragm. Experiments also this makes Jacobian transformations physiologically more plausible reliable.